NRMSE performance under burst model of missing values. Comparisons of NRMSE performances for KNN, LLS, GOKNN, GOLLS, knnHAI, and llsHAI in six datasets under burst model of missing values. The legends are the same as Figure 1. The HAIimpute methods are more robust than GOimpute methods in this case. The knnHAI method outperforms KNN and GOKNN, while llsHAI outperforms LLS and GOLLS in most cases.